Protein Devices May Increase Computer Speed And Memory - C&EN

Researchers at Syracuse University have developed a prototype protein- based optoelectronic device that may provide as much as 300 times more memory ...
0 downloads 0 Views 416KB Size
SCIENCE/TECHNOLOGY

Protein Devices May Increase Computer Speed And Memory yields a binary 1 output signal if and only if the two input signals are in the binary 1 state. In their prototype device, a green laser beam is used to initiate the bacteriorhodopsin photocycle in an ultrathin section of the cube. In effect, a "page" of molecules is photoactivated to initiate the photocycle. Two milliseconds later, before the molecules can Michael Freemantle, C&EN London relax back to their resting state, a twodimensional array of red lasers at right esearchers at Syracuse University angles to this beam is fired at the page. have developed a prototype proThis induces the photochemical branchtein-based optoelectronic device ing reaction to the "Q state" in the that may provide as much as 300 times branched photocycle in sets of molecules more memory than devices currently on the page targeted by the red lasers. used in computers. The device uses laser For computational purposes, these molbeams to write information into and ecules are converted to the binary 1 read information out of cubes containing state. Nontargeted molecules on the the protein bacteriorhodopsin. page return to the resting state along Bacteriorhodopsin is a light-harvestwith all the other molecules in the cube. ing protein found in a bacterium that Computationally, these molecules are all grows in salt marshes. When the proin the binary 0 state. Both states remain tein absorbs light, it undergoes a comstable for several years. plex photocycle resulting in dramatic changes in its optical and elec"Because you can move the tronic properties. page from one location to another inside the cube, you can rigorousThese unique photophysical Bacteriorhodopsin functions as ly select locations in the memory properties were discovered in computational gate to write data," says Birge. The the early 1970s. Since then, bacdata are read by activating the teriorhodopsin's potential for AND Input Output page with the green beam again. use in optoelectronic devices has gate Low-intensity red lasers are then excited the interest of scientists Green Protein(bR) fired at the page. Molecules in the around the world. A key breakpaging cycles but laser returns to binary 0 state absorb the light through came last year when resting state whereas those in the binary 1 the Syracuse researchers discov(returns to 0) state allow the light to pass ered by chance a photochemical through. The entire page of binabranching reaction in the prory output signals is imaged by a tein's photocycle. charge-injection device array. No absorption It is this branching reaction that and no change bR Data can be written in parallel provides for long-term, high-denRed in protein (0) at multiple locations on die illusity volumetric data storage, exdata(stays at 0) laser minated page. Similarly, the plains chemist Robert R. Birge, dimultiple bits of data on a page rector of the W. M. Keck Center Green can be read simultaneously. The for Molecular Electronics and repaging Protein cycles simultaneous manipulation of search director of the New York laser and then multiple sets of data is known as State Center for Advanced Techbranches to stable Q state Red parallel processing. Either senology in Computer Applications (converts 0 to 1 ) data quence takes about 10 millisec& Software Engineering at Syralaser onds, which is equivalent to cuse University. writing or reading data at a rate Birge described the serendipi-

• Prototype devices use bacterial protein memory cubes to store up to 300 times more information than current devices

R

tous breakthrough in a lecture on optoelectronic devices based on bacteriorhodopsin at a conference on "The Science of Complex Molecular Systems," held last month in Paris. The conference was Nature's third international conference in Europe. "One of the problems in making viable three-dimensional memories has always been that when you try to read or write information inside a volumetric solid, you cannot go back to the same location accurately, and you cannot rigorously exclude writing outside of the area that you are trying to write to," says Birge. The Syracuse discovery overcomes this problem. "We use the protein as an optical AND gate," explains Birge. An AND gate is one of five main types of logic gates used in building integrated circuits. Binary-coded input signals are fed into the gate. The binary 0 state corresponds to "off," and the binary 1 state corresponds to "on." An AND gate

{

{

{

24

MAY 22,1995 C&EN

of 10 megabytes per second for each page of memory. This speed is comparable to a slow semiconductor memory, according to Birge. The overall speed of the memo­ ry is directly proportional to the num­ ber of cubes operating in parallel. An eight-cube memory would operate at 80 million characters per second. Parallel processing is dependent on a "sequential one-photon" process, Birge says. The first photon, from the green laser, activates the photocycle, and the second photon, from the red laser, acti­ vates the branching reaction to form a stable photoproduct. "This sequential one-photon architecture completely eliminates unwanted photochemistry outside of the irradiated volume," he notes. In principle, an optical three-dimen­ sional memory can store approximately 1,000 times more information in the same size enclosure compared with a two-dimensional optical disc memory. Optical limitations and issues of reliabil­ ity lower this ratio, however. "A threedimensional memory stores roughly 300 times more information in the same volumetric enclosure," Birge ex­ plains. "This means you can make a memory that would store about three and a half gigabytes [3.5 billion bytes] on a card that would fit into a typical computer." The Syracuse team has bench-top models of the bioelectronic device op­ erating in its laboratory. These fall into the category of level I prototypes. Lev­ el II prototypes are miniaturized ver­ sions that can be put into a computer. "My prediction is that we will have working level II prototypes in three to four years," says Birge. The devices will need further im­ provement before they reach the level III prototype stage. At this stage, ex­ plains Birge, the devices will be ready to manufacture. "The field is moving very quickly, and it's well funded," he says. Birge's research is sponsored by the U.S. Air Force Rome Laboratory, the W. M. Keck Foundation, the Na­ tional Institutes of Health, and 12 com­ panies including four major U.S. com­ puter manufacturers. The aim is to make a memory card with the protein encapsulated in a poly­ mer. "The protein is very cheap for stor­ ing data compared with semiconduc­ tors," comments Birge. He notes that, at current prices, disks cost approximately

$1.00 per megabyte of memory. "Our goal is to make a 3.5 gigabyte card which would sell for about $600. That is a considerable savings, although I'm sure prices of disk drives will drop. However, I anticipate our card will be cost-effective to the user." One of the major attractions of using the bacteriorhodopsin in computer memories is its thermal and photo­ chemical stability. The protein occurs in a lipid matrix in the purple mem­ brane of a bacterium that grows in salt marshes where the salt concentration is around six times that of seawater. The bacterium, called Halobacterium halobium, is able to survive in the harsh conditions of high temperature and in­ tense light in the salt marsh by switch­ ing from respiration to photosynthesis when the need arises. The bacterium grows the purple membrane when oxygen concentrations become too low to sustain respiration. It absorbs energy from light by means of a chromophore embedded in the protein. When the chromophore absorbs light, the structure of the protein changes and the bacterium is thus able to pump a proton across the membrane. This gives rise to a chemical and osmotic potential that serves as an alternative source of energy. Bacteriorhodopsin can undergo more than 1 million photochemical cycles without damage. According to Birge, this value is much higher than most syn­ thetic photochromic materials. And, he notes, "Few semiconductor devices could survive a salt marsh." The Syracuse researchers grow the bacteria in simulated salt marsh condi­ tions in their laboratory. "Isolating the protein is essentially straightforward," comments Birge. The team uses geneti­ cally modified bacteria that produce 10 times as much of the protein as normal, he says. "We can create about 500 mg a week per 14-L vessel. That's a lot of material." Chemistry professor George M. Whitesides of Harvard University drew atten­ tion to the stability of the protein in his summary of the Paris conference. "Bacte­ riorhodopsin, in the hands of Birge, has proved to be much more robust and has more interesting properties than I think many people would ever have guessed," he comments. "It is a real example of how you can use proteins for something that requires real stability." Birge's lecture at the Paris conference

ο ο -c CL

it: •2 CO

Birge: development of hybrid technology

was one of four lectures presented in a section on molecular electronics. Mo­ lecular electronics, according to Birge, can be broadly defined as the encod­ ing, manipulation, and retrieval of infor­ mation at a molecular or macromolecular level. Several researchers comment­ ed on the limitations of semiconductor technology and the potential advantag­ es of using molecular electronics in computers. Allen J. Bard, Hackerman/Welch Re­ gents Chair in Chemistry at the Univer­ sity of Texas, Austin, tells C&EN: "I think, in silicon technology as practiced now, one will reach a stopping point based on the material capabilities. You can only do so much with lithography and with small wires." Michael C. Petty, codirector of the University of Durham Centre for Molec­ ular Electronics in England, reinforces the point. "There's a lot of interest in materials other than silicon and gallium arsenide for electronics because these are somewhat limited in their applications," he tells C&EN. "I don't think we're try­ ing to replace the silicon chip. But there are lots of areas where organic materials might fulfill a specific role. What Birge is doing is extremely exciting. Clearly, he's bringing together expertise from the background of physics, chemistry, and biology." Birge stresses that the development of a purely biomolecular computer is not a serious proposition. He sees the development of a hybrid technology based on a combination of biomolecules and semiconductors as far more MAY 22,1995 C&EN 25

SCIENCE/TECHNOLOGY likely. "Liquid crystal display technology is a prime example of a hybrid molecular-semiconductor technology that has achieved widespread success/' he says. One of the problems that has to be solved with protein memory cubes is the provision of an environment for the protein that will give it stability over decades rather than a few years. "Right now, our memories will store information for five years plus or minus a year without any trouble," says Birge. "It's going to take a lot of additional work not only using chemistry but also using molecular biology to make something that can store data on the order of 20 years." Birge considers protein memory cubes to be more environmentally friendly than semiconductors, however. "There are several steps in making a semiconductor device, such as cleansing and etching, that involve a major amount of pollution," observes Birge. "The preparation and isolation of the protein is environmentally safe, and you can even eat the protein. It is nutritious, but not delicious." The Syracuse team is also working on other types of memory applications for bacteriorhodopsin. One of these is the use of the protein to build neural networks that mimic the learning-byassociation capabilities of the brain. This type of memory is known as associative. An associative memory receives an image, or block of data, and then searches the entire memory for a data block that matches the input. The memory then returns the matching or most closely matching block as output. "You feed in a small amount of data and the memory finds that location in memory that has the best overlap with the data you fed in. It then feeds back not only that piece of information but all the associated information," explains Birge. Birge's laboratory has developed an associative-memory device that relies on the holographic properties of thin films of bacteriorhodopsin. "The device takes advantage of the fact that when the protein is activated with light, it changes its refractive index dramatically," he says. "During this refractive index change, it can function as a holographic recording material." Holograms allow multiple images to be stored in the same memory segment. This enables large sets of data to be analyzed simultaneously. • 26

MAY 22,1995 C&EN

New journal forgoes traditional peer review Rustum Roy doesn't think much of peer review as it is traditionally practiced. For years, he has complained that it slows the dissemination of research findings and is biased against truly innovative results. He favors peer review of past performance instead. Almost three years ago, Roy, a prominent materials scientist at Pennsylvania State University, caused a stir by announcing a research advance in diamond synthesis at a press conference before any journal had even had a chance to review the work (C&EN, Oct. 26, 1992, page 5). And now he's fired another salvo in his campaign against traditional peer review by preparing to launch a materials science journal that will rapidly publish "new, unexpected, and innovative" results without the delays caused by reviewers. Other materials researchers contacted by C&EN do not regard the peer review system as an outdated relic of the past— as Roy does—and they are not enthused by a new journal that throws traditional peer review out the window. At peer-reviewed journals, the editors decide whether to publish a paper based largely on the comments of scientists not directly connected with the research who review the manuscript in

Roy: innovative results without delays

detail. Roy's new journal, Innovations in Materials Research (IMR), will examine the author's past publications track record "instead of nitpicking on the details of [the] submitted manuscript," Roy explains in the journal's publicity release. Authors who have published 50 papers or patents (fewer for those working in industry) will have to submit their bibliographies to establish their track record. This process, which Roy calls "super peer review," places full responsibility for the paper's content on the authors and "uses the authors' published record as guarantee." Researchers who haven't established the necessary track record may submit their paper via a colleague who has and who agrees to sponsor them. Alternatively, younger or less prolific authors may submit their bibliography and indicate what they regard as the most significant and innovative papers. The other key feature of IMR, according to Roy, is its focus on the new and unexpected. The journal's brochure notes that "dozens of journals are available for publishing the enormous amounts of good incremental science done in every subfield of materials research. But [papers reporting major] and unexpected advances are buried in a great deal of other material. Moreover, it is precisely these papers that encounter the greatest resistance in the traditional 'peer review' system. [IMR] will focus on such genuinely innovative papers." To help the IMR editors judge the suitability of a paper, the manuscript must point out its key innovations and how the research differs from or advances previous work. This section of the paper also must indicate the extent to which the author searched the literature to find relevant previous work in the field. Roy tells C&EN that the scientific literature is littered with erroneous reports, most of which are not corrected, and only some of which have been retracted. When retractions are published, they often are not titled in a way that allows them to be picked up in computerized literature searches, he points out. And some journals have even refused to publish retractions. IMR hopes to avoid these problems, while at the same time encouraging discussion and criticism. To this end,